A category entry point (CEP) is a situation that makes someone think of a product category in the first place: the moment, need, place, mood or constraint that starts the search. Brands do not compete for attention in general. They compete to be the one that comes to mind in a specific situation, and the set of situations a brand is linked to predicts its share better than how much people say they like it.
This matters for measurement in a way that is easy to miss. Every method in the marketing measurement stack - attribution, mix modeling, incrementality testing - starts from your spend and works forward to an outcome. A category entry point runs the other way. It starts in the buyer's life, before any of your media reaches them, and it determines whether your brand was even a candidate. The causal variable sits outside the model by construction. See MMM vs attribution vs incrementality for what those methods can and cannot prove; this article is about the demand they are all downstream of.
The number that reframes the problem
Professor John Dawes of the Ehrenberg-Bass Institute put a figure on how much demand is not currently in play. Writing with LinkedIn's B2B Institute in May 2021, he estimated that up to 95 percent of business clients are not in the market for many goods and services at any one time. Roughly 20 percent of business buyers are in the market over the course of an entire year, and something like 5 percent in a given quarter.
The strategic consequence is direct, and Dawes states it plainly: advertising mostly reaches buyers who are not going to buy anytime soon, so it "mainly works by building and refreshing memory links to the brand", and those memory links "activate when buyers do come into the market."
If that is how advertising works, then the thing you are building is a set of associations between your brand and the situations that will eventually occur. Those situations are the category entry points. Everything else - the creative, the channel, the frequency - is machinery for attaching your brand to them.
The 95-5 split was derived in a B2B context and the exact ratio varies enormously by category (people buy yogurt more often than enterprise software). The structural point survives the variation: at any given moment most of your future buyers are not shopping, are not searchable as in-market intent, and are not in any conversion dataset you own.
Why your measurement stack cannot see this
Three well-documented results explain why demand origination is invisible to outcome-based measurement.
The ad often finds demand it did not create. When eBay halted brand-keyword paid search on MSN in March 2012, Blake, Nosko and Tadelis found that 99.5 percent of the forgone paid click traffic was immediately captured by natural search (Econometrica 83(1):155-174, 2015). The purchase intent already existed. The ad was standing where the traffic was going to pass anyway. Attribution recorded it as a cause.
Even rich outcome data does not recover mechanism. Gordon, Zettelmeyer, Bhargava and Chapsky compared randomized and observational estimates across 15 Facebook experiments with 500 million user-experiment observations and found the observational estimates off by a factor of three or more in half the studies (Marketing Science 38(2):193-225, 2019). Better features did not reliably help. Nothing in that data records what the buyer was trying to do.
The signal is tiny relative to the noise. Lewis and Rao showed that individual-level sales have a standard deviation around ten times the mean, so a highly profitable campaign implies a model fit on the order of an R-squared of 0.0000054 (Quarterly Journal of Economics 130(4):1941-1973, 2015). You are not going to reverse-engineer buying situations from that residual.
So the situation that started the purchase never appears in the purchase data. It has to be collected directly, from people, about specific occasions.
What a category entry point actually looks like
The academic foundation is Jenni Romaniuk's work on mental availability at the Ehrenberg-Bass Institute, formalized in "Modeling mental market share" (Journal of Business Research 66(2):188-195, 2013) and developed for practitioners in How Brands Grow Part 2 (Romaniuk and Sharp, 2016) and Better Brand Health (Romaniuk, 2023). Mental market share is the share of a brand's links to category entry points relative to competitors - a measure of how much of the category's situational territory your brand occupies in memory.
Romaniuk's 7W framework is the standard elicitation scaffold. A well-formed CEP usually answers one or more of:
| W | Question it answers | Example (coffee) |
|---|---|---|
| Why | What is the underlying motivation | "I need to stay awake for a late deadline" |
| When | What time or trigger point | "First thing, before anyone talks to me" |
| Where | What physical or digital location | "On the train platform" |
| While | What co-occurring activity | "While I am reading the news" |
| With whom | Who else is present | "When a colleague suggests a break" |
| With what | What it is consumed alongside | "With something sweet after lunch" |
| hoW feeling | What emotional state | "When I am flat and need to reset" |
Two properties separate a real CEP from a marketing slogan. First, it is stated from the buyer's point of view, in the buyer's language, and it describes a situation rather than an attribute. "Premium quality" is not a CEP. "When I want to make a good impression on a guest" is. Second, it is a retrieval cue: it is the thing present in the moment that could make a brand name surface.
CEPs vs jobs-to-be-done vs trigger events
These three frameworks overlap enough to be confused constantly, and they are not interchangeable.
| Category entry points | Jobs to be done | Trigger events | |
|---|---|---|---|
| Core object | A situation that cues the category | A goal the customer is hiring a product to achieve | A discrete event that starts a buying process |
| Time horizon | Recurring, often habitual | Stable across occasions | A single moment in a specific deal |
| Primary use | Building mental availability and media planning | Product design, positioning, segmentation | Sales timing and demand capture |
| Measured as | Share of situations the brand is linked to, across a population | Progress, struggle and switching forces | Presence or absence of a signal |
| Typical unit | The occasion | The customer | The account or deal |
| Sample needed | Large, quantified | Small, deep | Operational, ongoing |
They are complements. A jobs-to-be-done interview explains why the job exists and what competes with you for it; see the jobs-to-be-done framework and switch interviews for that method. Trigger event interviews explain what starts a specific purchase process. CEP research explains which recurring situations your brand does and does not own across the whole category - and it is the only one of the three that produces a comparable population-level share number.
How to run CEP research in two phases
The practical design that has become standard is two-phase, and it maps precisely onto the two things research has to do here: discover the situations, then size them.
Phase 1: elicit the situations, qualitatively and in the buyer's words
You cannot write a good CEP list from a workshop. Internal teams generate attributes, not occasions, and they generate the occasions the brand already serves. The situations you are missing are, by definition, the ones nobody in the building thinks about.
The method is occasion-based recall: ask a buyer about the last several times they used or bought in the category, and walk each one through the 7Ws. Not "what do you look for in a coffee" but "tell me about the last time you bought a coffee that you had not planned to buy - where were you, what were you doing, what was going on that day." Then probe. The follow-up is where the real cue emerges, because the first answer is almost always a rationalization.
Two design cautions matter here:
- Recall degrades fast and reconstructs itself. People do not retrieve occasions, they infer them from a general impression. Anchor on specific recent instances rather than typical behavior; see recall bias for how to structure this.
- The question order will create the answer if you let it. Asking about your brand first contaminates every situation that follows. Elicit situations before you introduce any brand names at all.
Phase 2: quantify frequency and mental market share
Phase 1 yields a long list of candidate situations. Phase 2 turns it into a map: for each situation, how often does it occur, how many category buyers experience it, and which brands come to mind when it does. That produces the two numbers you act on - the size of the situation, and your share of it.
The output is a prioritized grid. A situation that is large and where you are weak is a growth opportunity. A situation that is large and where you are strong is an asset to defend. A situation that is small and where you are strong is often where the entire marketing budget has been going, because it is the one the brand team is most comfortable with.
Why this used to be a six-figure project
Phase 1 needs qualitative depth from a lot of people - a handful of interviews will not surface a category's situational territory. Phase 2 needs structured, comparable data across a representative sample. Traditionally those are two different studies, two different vendors, two different timelines, and a translation loss in between: the qualitative team hands over a list, the quant team turns it into scale items, and the buyer's actual language is gone by the time anyone measures anything.
That is why most brands skip it and run an attribute tracker instead. Attribute trackers are cheap, familiar, and measure something that does not predict much.
How Koji makes CEP research a single study
Koji runs AI-moderated voice interviews at survey scale, which collapses the two phases into one instrument.
- Phase 1 at sample size. The AI consultant conducts occasion-based interviews with hundreds of category buyers in parallel, probing each situation with follow-up questions the way a good moderator would. You get the raw language of hundreds of buying occasions instead of the twelve you could afford to moderate by hand.
- Automatic thematic analysis. Koji clusters the occasions into candidate CEPs and surfaces the ones your team never articulated, with supporting quotes attached. No manual coding pass, no six-week synthesis.
- Phase 2 in the same study. Six structured question types run alongside the conversation:
open_endedfor the occasion narrative,scalefor frequency and importance,single_choiceandmultiple_choicefor situation membership and brand linkage,rankingfor prioritizing situations against each other, andyes_nofor clean qualification. The structured questions guide covers how each type is analyzed. That gives you a sizeable, comparable mental market share number from the same respondents who told you the stories. - No moderator bias. Every respondent is asked the same core questions, so the situational data is genuinely comparable across the sample rather than shaped by who was in the room.
- Hours, not weeks. Which means CEP research can be refreshed when the category changes rather than once every three years.
Related reading in the docs: Koji for marketing teams, the product marketing research playbook, customer journey interviews, competitive intelligence surveys, and AI research for CPG.
Five mistakes that kill CEP projects
- Writing the CEP list internally. You will produce your own positioning back to yourself. Elicit from buyers, always.
- Confusing attributes with situations. "Trusted", "premium" and "innovative" are not entry points. If it cannot complete the sentence "when I...", it is not a CEP.
- Measuring only your own buyers. Your customers are, by construction, the people whose situations you already serve. The growth is in the situations you lose, which means interviewing category buyers, not your CRM.
- Stopping at discovery. A qualitative list with no sizing produces a debate, not a decision. Phase 2 is what makes it actionable.
- Treating it as a one-off. Situations change when the category changes. A CEP map from 2022 does not describe how people buy in 2026.
Want to find the buying situations your brand does not own? Start a study with Koji - occasion-based interviews with hundreds of category buyers, thematically analyzed and sized, in hours rather than months.
Frequently Asked Questions
What is a category entry point in simple terms?
A category entry point is a situation that makes someone think of a product category. It is a retrieval cue: the moment, need, place, activity, company or mood that causes a category and the brands in it to come to mind. "When I need to stay awake for a late deadline" is a category entry point for coffee. "High quality beans" is not - it is an attribute, and attributes do not start purchases.
How many category entry points should I measure?
Enough to cover the category's territory without exhausting respondents. Discovery typically surfaces dozens of candidate situations; most trackers carry a prioritized subset of roughly 15 to 25 into ongoing measurement, chosen by how often the situation occurs and how much of the category volume it represents. The right number depends on category complexity - a sizing phase, rather than a rule of thumb, is what should decide it.
How are CEPs different from jobs to be done?
They answer different questions at different units of analysis. Jobs to be done describes the progress a customer is trying to make and what competes for that job; it is deep, customer-level, and used for product and positioning decisions. Category entry points describe the recurring situations that cue the category, measured at the occasion level across a population, and used to build mental availability and plan media. They are complements, not substitutes.
Why can attribution or mix modeling not identify category entry points?
Because both start from your spend and work forward to an outcome, while a category entry point originates in the buyer's life before any media reaches them. The situation never appears in impression, click or transaction data. The eBay experiment illustrates the consequence: 99.5 percent of paid brand-search clicks were recovered by natural search when the ads were switched off, meaning the demand existed independently and the ad merely intercepted it. Outcome data cannot distinguish creating demand from meeting it.
What is mental market share?
Mental market share is a brand's share of links to category entry points relative to competitors - how much of the category's situational territory the brand occupies in buyers' memory. It was formalized in Jenni Romaniuk's "Modeling mental market share" (Journal of Business Research, 66(2):188-195, 2013). Unlike awareness or favorability, it is measured situation by situation, which makes it diagnostic: you learn not just that you are weaker, but exactly which occasions you are absent from.
Can I run CEP research with a survey alone?
Only phase two. A survey can efficiently size a list of situations and measure brand linkage, but it cannot generate the list, because you can only ask about situations you already thought of. Discovery requires open-ended, probed conversation about specific recent occasions in the buyer's own words. Running the survey without the discovery phase produces a confident measurement of the wrong situations - which is the most common failure mode in brand tracking generally.
